Hi All,
Noticed this from yesterday on the infra mailing list. I don't think
Pier's solution will do what we want but it at least gives insight
into why ASF does not want Confluence running on their servers.
I am going to ask about trying to set up confluence on a zone...
P.S. Sorry for the forward instead of a url pointer, I could not find
an archive of [EMAIL PROTECTED]
TTFN,
Bill Dudney
MyFaces - http://myfaces.apache.org
Cayenne - http://incubator.apache.org/projects/cayenne.html
Begin forwarded message:
From: Pier Fumagalli <[EMAIL PROTECTED]>
Date: March 18, 2006 6:22:43 PM MST
To: Apache Infrastructure <[EMAIL PROTECTED]>
Subject: Confluence take 2
DO NOT SHOOT THE MESSENGER! :-) Please, read below before you flame
the bejesus out of me!
Folks, I can understand why you people don't want to run Confluence
on the ASF servers (yep, it's a hog, as big as Jira), but a lot of
people want it, so, rather than ranting about all its benefits,
some months ago I went to the drawing board and started trying to
fix the problem...
I think that what we want is a wiki that isn't going to kill the
entire server when Slashdot points to it (or at least, that's what
I feel is the main concern), and that even if it's down, well, at
least the content and the doccos are up and visible.
I've used Confluence for years now and I can understand why lots of
people want to use it, but having to administer my servers, well, I
feel your pain! When I started my "fix" I had a chat with Jeff and
I showed him what I came out with, and this weekend I met Mads, and
he re-confirmed me that I was going in the right direction...
So, what's this all about? Well, I wrote a small confluence plugin
<http://could.it/confluence-autoexport-plugin.html> called
"AutoExport". What it does is that it listens to update/create/
remove events within confluence and exports (or better
synchronizes) the content in confluence with a set of files on a
disk, so that they can be served by Apache (it rewrites links in
the HTML, exports attachments, pre-processes thumbnails,
everything)...
Basically if you look at "http://could.it/";, it's all backed by
Confluence itself, but it's served directly by Apache from the
disk. Confluence uses 64 megs of ram (the JVM is limited) and
unless someone is editing something, it's doing absolutely nothing
(the only "extra" thing it does, is that when a page is saved, a
file is written on the disk using Velocity).
And it's easy to install/configure/use... I've made my "sales
pitch" for the Atlassian Codegeist <http://confluence.atlassian.com/
display/CODEGEIST/AutoExport+for+Confluence> (you can see some
screenshot, read some of the features).
Now, I want to be bold and install it on for the ASF. I can open up
my server if people want but I only have a 2 megs synchronous pipe
(it's not a crappy ADSL, but still, it's only 2 Megs). Otherwise,
if someone is nice enough to give me a root password, I was
thinking to install it directly on one of the ASF's servers (I was
thinking Ajax, Mads mentioned Brutus this weekend).
My idea is to lock-down the entire confluence wiki to those people
with the right to edit the documents (basically the committers on
the projects willing to use Confluence) and to allow anonymous
access _only_ through the auto-exported content (that's what I do
on http://Could.IT , you the only thing you can see of Confluence
itself without a password is the login page http://could.it/wiki/ )
So, what do you think? Shall we give it a try? I've been running
this for roughly two months now, and it works nicely. I'm coming
out with it only now because the Atlassian CodeGeist forced me to
clean up the code and to make all those usability improvements that
were lacking on the old version (template editing through the
browser, nice configuration integrated with Confluence, blablabla).
Pier
